Home Services Why Us Comparison FAQ Process Careers Blog Glossary Contact us
Illustration of A/B testing: two page versions, A and B, side by side

⚡ In short
What is A/B testing? A/B testing, also called split testing, shows one version of a page to half of your visitors and another version to the other half, then counts which one gets more of the action you want: leads, calls, purchases. It lets you change a site based on numbers.

What it is

A/B testing is an experiment run on live visitors. You take one page and make two versions, A and B, that differ in one detail: a headline, the button text, a picture or the order of blocks. The system splits traffic in half, so some people see A and others see B.

After a few weeks you compare which version leads to the target action more often. The version with the better result wins, even if the team likes the other one more.

A/B testing examples: what to test on a website

  • Headline and first screen. Two different promises on the same page get different reactions, and you can't guess in advance which one will win.
  • Button. The text ("Order" or "Get a quote"), the color and the position on the page.
  • Form. How many fields to ask for and in what order: every extra field loses some people.
  • Image or video. A team photo versus an illustration, a video versus a still frame.

Want to find out what works on your site?

We'll suggest what to test first and help prepare the page.

Get a consultation Telegram info@lulo.tech

How to run an A/B test

Start with a hypothesis: what you are changing and why you expect more leads. Change one thing at a time, otherwise you won't know what worked. Let the test run for a few weeks so enough visitors arrive. Look at conversion, the share of people who took the target action. Clicks alone prove nothing.

An example of such a test

Imagine a page that offers a free quote. In version A the button says "Submit", in version B it says "Get a quote in 5 minutes". The rest of the page is identical. Visitors are split evenly, and over three weeks 18 of 600 people fill in the form in A and 27 of 600 in B. Conversion rises from 3% to 4.5%, so B stays. The numbers are made up, but the scheme is exactly this.

How to read the result

A difference of a few leads does not prove that one version is better: on a small sample chance can easily produce it. Stop the test once enough visitors have arrived and the lead holds for several days. If you stop it the moment one version pulls ahead, the result may be random. Sometimes the versions barely differ. That is a result too: the change does not matter, and the gains are somewhere else. Keep the winner and build the next test on top of it.

A/B testing and usability testing: what is the difference

Usability testing shows why people struggle: you watch them use the site and listen to their comments. UX testing in the broader sense also gives this kind of qualitative answer. An A/B test answers a different question: which of two versions works better on real traffic. They work well together: first a website audit and observation reveal the problem, then an A/B test checks whether the fix helped.

What you need for a test

The main thing is traffic. If only a few dozen people visit a page per week, the result will be random, and it is better to fix the obvious problems first. You also need analytics that counts the target actions, and a fast site: slow loading spoils both versions, which our site speed case study shows. On most A/B testing website projects, the first page to test is the landing page: one page, one goal, changes show up quickly. If you do not have a site yet, tests can be planned in advance when a turnkey website is built.

Where else A/B tests are used

A/B tests are used beyond websites. In email campaigns they compare subject lines, in advertising two texts or two images, in apps the position of buttons. The logic is the same everywhere: two versions, equal conditions, one metric.

Common mistakes

Tests are most often stopped too early, when one version pulls ahead and visitors are still few. The second mistake is changing the headline, the button and the picture at once, after which nobody knows what worked. Then there is testing small details on a page that almost nobody visits. And finally, not writing down the hypothesis: a month later no one remembers what was being checked. You can also compare lead magnet formats this way: what people take more willingly, a checklist or a template. What is worth checking usually comes from UX/UI design, which shows where people struggle.